<![CDATA[Small businesses often handle customer questions through manual replies via chat applications or phone calls, causing repetitive work, delayed responses, and inconsistent information delivery. This study proposes a web-based FAQ chatbot that answers user questions by performing semantic search over an Indonesian FAQ knowledge base and ranking the most relevant response. The chatbot applies a lightweight information retrieval approach using TF-IDF vectorization and cosine similarity to compute the relevance score between the user query and FAQ entries (question and tags). The system then selects the top-ranked FAQ entry and returns its associated answer, meaning the semantic matching is performed at the question-to-question level, not directly between questions and answers. The top results are ranked, and the chatbot returns the best answer along with a confidence score and the top three candidate questions to increase transparency. If the score is below a predefined threshold, the system provides a fallback response and suggests related topics rather than forcing an incorrect answer. The system is implemented as a PHP–MySQL web application with an administrator dashboard that supports secure login, FAQ CRUD management, chat logging, and usage analytics. Functional verification is conducted using black-box testing across main modules, including authentication, FAQ management, chatbot interaction, logging, and analytics dashboards. The expected contribution of this work is a practical and low-cost chatbot solution that can be deployed by small businesses to reduce repetitive customer service workload, accelerate response time, and provide measurable service insights through log-based analytics. Future improvements include expanding the knowledge base, enhancing Indonesian text normalization, and adopting embedding-based retrieval for better semantic matching.]]>
Copyrights © 2026
